Unibase: organise saved context and move it between AI chats

Decide what context belongs in the memory

Thanks to Unibase for sponsoring this video.

The demonstration imports conversations and saves webpages or selections. A complete chat history can contain irrelevant, outdated or sensitive material.

1. Install the extension through the product’s documented route.
2. Choose individual conversations or the intended recent-chat sync scope.
3. Save a whole page only when the full context is useful; otherwise keep the relevant selection.
4. Review the product’s current storage and privacy documentation before including sensitive material.

Make saved material retrievable

Tags and stars organise the collection; searching across it should bring back the relevant source rather than a vague recollection.

1. Tag the subject, project and date where those distinctions matter.
2. Star durable context you actually reuse.
3. Use Ask/search to find one known item and inspect the returned source.
4. Keep stale decisions distinguishable from current instructions.

Send a deliberate context package

The recording moves useful context between ChatGPT and Claude. Finding a memory does not mean every saved item needs to be shared.

1. Select the exact material needed for the new task.
2. Inspect the content and destination before sending it to another AI.
3. Ask the receiving assistant to distinguish source facts from assumptions.
4. Verify the resulting answer against the original saved material, especially when it contains dates or decisions.

The takeaway

Shared memory reduces repeated context transfer when the saved material is relevant and controlled. Do not assume that local storage and optional encrypted sync have identical data flows.